The cost of measurement data errors and how to prevent them

What a data error can actually cause

The impact of a data error depends on where it ends up. The same missing value can mean an incorrect bill, a missed fault, or a failed audit. Here is how these scenarios play out in practice.

Incorrect cost allocation

In a residential building with 40 tenants, a 5% error in heat measurement can mean some tenants overpay, whilst others pay less than they should. The result is complaints, administrative burden, and, in the worst case, legal disputes. In commercial buildings with multiple tenants, the situation is even more complex. Energy cost allocation is part of lease agreements, and any discrepancy can be legally challenged.

Decisions based on data that does not lie but misleads

If some meters stop sending data and the system fails to detect this, consumption may be assessed as lower than it actually is. The operator then incorrectly evaluates an efficiency measure as successful and invests further resources into it. In reality, they are making decisions based on incomplete data. The impact is felt not only in billing but also in long-term strategy. The same principle works in reverse: an unusual drop or spike in consumption can be the first signal of a water leak, a faulty pump, or overheating equipment, long before anyone notices physically. Without automated anomaly detection, these signals are simply missed.

Reporting and audit problems

ESG reporting, ISO 50001, EED. All of these require verifiable, consistent, and auditable consumption data. Missing values, gaps in time series, or discrepancies between data sources are precisely what an auditor will flag first. In practice, this can mean having to supplement data retrospectively or repeat the entire reporting process, which is costly in both time and money.

Hidden costs that are difficult to quantify

These include the time technicians and analysts must spend manually correcting data, tracking down missing values, or explaining inconsistencies to clients. These costs are difficult to measure but significantly affect the system’s total cost of ownership over time.

Where errors originate

Data errors typically have one of the following sources:

• ‍Communication failure: the meter works, but data does not reach the system. This may be a temporary network outage, a weak signal, or a depleted converter battery.‍
• Incorrect configuration: the device sends data, but in the wrong format, with an incorrect unit, or with the wrong conversion coefficient. This is a very common installation error that may not become apparent for weeks.‍
• Meter or converter fault: a physical defect that can generate zero values, constant values, or values outside the valid range.‍
• Data processing error: data arrives correctly, but is processed incorrectly on the server or in the application. For example, a value being overwritten, an incorrect average calculated, or data lost during a system migration.

‍

The costs of unreliable data, from legal disputes to missed leaks, are high, but they are not inevitable. Most data failures are predictable results of a "fragile" system design. To move from reactive firefighting to proactive management, the infrastructure must be built with layers of redundancy and self-awareness. Prevention isn't about human oversight; it’s about a system that validates itself at every step. A truly resilient architecture is defined by these core capabilities:

On-device data archiving

Communication outages are inevitable. Devices that store data locally and send it later ensure no records are lost. ACRIOS converters retain data for up to 40 days, preventing loss even during prolonged outages. Without this, communication failures create data gaps that require estimation or acknowledgement as missing.

Data validation at source

A high-quality IoT converter should perform basic data checks before sending data, such as detecting values outside physical limits, monitoring power usage, and reporting error codes. The earlier anomalies are caught, the less they affect downstream systems.

Diagnostics, not just readings

A good metering system sends both readings and metadata, such as battery status, signal quality, and error codes, to indicate whether devices are working correctly. Devices that report only measured values are nearly useless for troubleshooting.

Notifications on outages and anomalies

The system should alert you when data stops or values fall outside the expected range, via email, SMS, or app notifications. What matters is timely, clear alerts that specify the device and problem. Passive monitoring with monthly reviews is not sufficient.

Anomaly detection

An unusual spike in water consumption overnight, a temperature value that remains constant for 48 hours, or a sudden drop in heat consumption during winter. These are signals that an automated system can capture and flag for manual review. A person simply cannot spot such details across dozens or hundreds of data points.

Data integrity during transmission

ACRIOS converters forward data as a standard M-Bus frame with built-in integrity controls. For NB-IoT, we support DTLS for encrypted, intact data transmission.

Open formats and standardised communication

Data in proprietary or poorly documented formats is a source of errors at every integration point. JSON, CSV, MQTT, HTTP POST: open and well-documented formats that ensure data can be easily processed, audited, and transferred to another system without risk of loss or distortion.

How to design a system that is resilient and auditable

The starting point is hardware with proper diagnostics. A converter that reports battery status, signal quality, error codes, and the timestamp of the last successful transmission gives you immediate visibility when something goes wrong, rather than leaving you to discover the problem during the next manual review.

Once the right hardware is in place, alerts matter more than dashboards. A system that actively notifies you when data stops arriving or when values fall outside the expected range lets you act on problems in hours rather than weeks. Paired with data stored at multiple levels, locally on the device (up to 40 days), on the server, and in a backup, you have a safety net that covers outages at any single point.

The layer that is most often overlooked is configuration documentation. Incorrect configuration is one of the most common causes of data errors and one of the most difficult to trace retrospectively. Keeping clear logs of settings and changes will save significant time during troubleshooting.

Finally, automation should handle the routine. The goal is not to remove people from the process entirely, but to free them from checking numbers so they can focus on the broader context that only a person can spot.

Summary

Data errors in metering systems are not inevitable. They are largely a consequence of the system's design. A device that archives, validates, and reports anomalies in data is the foundation of an infrastructure that operates reliably without daily supervision. If you are designing or reconsidering your metering system, start with a simple question: what happens if one device stops communicating for three days? If the answer is not "the system alerts me automatically and the data is backfilled once the connection is restored", it is time to reconsider the architecture.